Qiu Shuang, He Feng, Tang Jiabei, Xu Jiapeng, Zhang Lixin, Zhao Xin, Qi Hongzhi, Zhou Peng, Cheng Xiaoman, Wan Baikun, Ming Dong
Annu Int Conf IEEE Eng Med Biol Soc. 2014;2014:2561-4. doi: 10.1109/EMBC.2014.6944145.
Functional electrical stimulation (FES) could restore motor functions for individuals with spinal cord injury (SCI). By applying electric current pulses, FES system could produce muscle contractions, generate joint torques, and thus, achieve joint movements automatically. Since the muscle system is highly nonlinear and time-varying, feedback control is quite necessary for precision control of the preset action. In the present study, we applied two methods (Proportional Integral Derivative (PID) controller based on Back Propagation (BP) neural network and that based on Genetic Algorithm (GA)), to control the knee joint angle for the FES system, while the traditional Ziegler-Nichols method was used in the control group for comparison. They were tested using a muscle model of the quadriceps. The results showed that intelligent algorithm tuning PID controller displayed superior performance than classic Ziegler-Nichols method with constant parameters. More particularly, PID controller tuned by BP neural network was superior on controlling precision to make the feedback signal track the desired trajectory whose error was less than 1.2°±0.16°, while GA-PID controller, seeking the optimal parameters from multipoint simultaneity, resulted in shortened delay in the response. Both strategies showed promise in application of intelligent algorithm tuning PID methods in FES system.
功能性电刺激（FES）可为脊髓损伤（SCI）患者恢复运动功能。通过施加电流脉冲，FES系统可引起肌肉收缩，产生关节扭矩，从而自动实现关节运动。由于肌肉系统具有高度非线性和时变性，因此反馈控制对于预设动作的精确控制非常必要。在本研究中，我们应用了两种方法（基于反向传播（BP）神经网络的比例积分微分（PID）控制器和基于遗传算法（GA）的PID控制器）来控制FES系统的膝关节角度，而对照组则使用传统的齐格勒-尼科尔斯方法进行比较。使用股四头肌的肌肉模型对它们进行了测试。结果表明，智能算法调整的PID控制器表现出比具有恒定参数的经典齐格勒-尼科尔斯方法更优越的性能。更具体地说，由BP神经网络调整的PID控制器在控制精度方面表现更优，能使反馈信号跟踪期望轨迹，误差小于1.2°±0.16°，而GA-PID控制器通过多点同时寻找最优参数，减少了响应延迟。两种策略在FES系统中应用智能算法调整PID方法方面都显示出了前景。
